How Did the Universe Start?

I’m on record as predicting that we’ll understand what happened at the Big Bang within fifty years. Not just the “Big Bang model” — the paradigm of a nearly-homogeneous universe expanding from an early hot, dense, state, which has been established beyond reasonable doubt — but the Bang itself, that moment at the very beginning. So now is as good a time as any to contemplate what we already think we do and do not understand. (Also, I’ll be talking about it Saturday night on Coast to Coast AM, so it’s good practice.)

There is something of a paradox in the way that cosmologists traditionally talk about the Big Bang. They will go to great effort to explain how the Bang was the beginning of space and time, that there is no “before” or “outside,” and that the universe was (conceivably) infinitely big the very moment it came into existence, so that the pasts of distant points in our current universe are strictly non-overlapping. All of which, of course, is pure moonshine. When they choose to be more careful, these cosmologists might say “Of course we don’t know for sure, but…” Which is true, but it’s stronger than that: the truth is, we have no good reasons to believe that those statements are actually true, and some pretty good reasons to doubt them.

I’m not saying anything avant-garde here. Just pointing out that all of these traditional statements about the Big Bang are made within the framework of classical general relativity, and we know that this framework isn’t right. Classical GR convincingly predicts the existence of singularities, and our universe seems to satisfy the appropriate conditions to imply that there is a singularity in our past. But singularities are just signs that the theory is breaking down, and has to be replaced by something better. The obvious choice for “something better” is a sensible theory of quantum gravity; but even if novel classical effects kick in to get rid of the purported singularity, we know that something must be going on other than the straightforward GR story.

There are two tacks you can take here. You can be specific, by offering a particular model of what might replace the purported singularity. Or you can be general, trying to reason via broad principles to argue about what kinds of scenarios might ultimately make sense.

Many scenarios have been put forward among the “specific” category. We have of course the “quantum cosmology” program, that tries to write down a wavefunction of the universe; the classic example is the paper by Hartle and Hawking. There have been many others, including recent investigations within loop quantum gravity. Although this program has led to some intriguing results, the silent majority or physicists seems to believe that there are too many unanswered questions about quantum gravity to take seriously any sort of head-on assault on this problem. There are conceptual puzzles: at what point does spacetime make the transition from quantum to classical? And there are technical issues: do we really think we can accurately model the universe with only a handful of degrees of freedom, crossing our fingers and hoping that unknown ultraviolet effects don’t completely change the picture? It’s certainly worth pursuing, but very few people (who are not zero-gravity tourists) think that we already understand the basic features of the wavefunction of the universe.

At a slightly less ambitious level (although still pretty darn ambitious, as things go), we have attempts to “smooth out” the singularity in some semi-classical way. Aguirre and Gratton have presented a proof by construction that such a universe is conceivable; essentially, they demonstrate how to take an inflating spacetime, cut it near the beginning, and glue it to an identical spacetime that is expanding the opposite direction of time. This can either be thought of as a universe in which the arrow of time reverses at some special midpoint, or (by identifying events on opposite sides of the cut) as a one-way spacetime with no beginning boundary. In a similar spirit, Gott and Li suggest that the universe could “create itself,” springing to life out of an endless loop of closed timelike curves. More colorfully, “an inflationary universe gives rise to baby universes, one of which turns out to be itself.”

And of course, you know that there are going to be ideas based on string theory. For a long time Veneziano and collaborators have been studying what they dub the pre-Big-Bang scenario. This takes advantage of the scale-factor duality of the stringy cosmological field equations: for every cosmological solution with a certain scale factor, there is another one with the inverse scale factor, where certain fields are evolving in the opposite direction. Taken literally, this means that very early times, when the scale factor is nominally small, are equivalent to very late times, when the scale factor is large! I’m skeptical that this duality survives to low-energy physics, but the early universe is at high energy, so maybe that’s irrelevant. A related set of ideas have been advanced by Steinhardt, Turok, and collaborators, first as the ekpyrotic scenario and later as the cyclic universe scenario. Both take advantage of branes and extra dimensions to try to follow cosmological evolution right through the purported Big Bang singularity; in the ekpyrotic case, there is a unique turnaround point, whereas in the cyclic case there are an infinite number of bounces stretching endlessly into the past and the future.

Personally, I think that the looming flaw in all of these ideas is that they take the homogeneity and isotropy of our universe too seriously. Our observable patch of space is pretty uniform on large scales, it’s true. But to simply extrapolate that smoothness infinitely far beyond what we can observe is completely unwarranted by the data. It might be true, but it might equally well be hopelessly parochial. We should certainly entertain the possibility that our observable patch is dramatically unrepresentative of the entire universe, and see where that leads us.


Inflation makes it plausible that our local conditions don’t stretch across the entire universe. In Alan Guth’s original scenario, inflation represented a temporary period in which the early universe was dominated by false-vacuum energy, which then went through a phase transition to convert to ordinary matter and radiation. But it was eventually realized that inflation could be eternal — unavoidable quantum fluctuations could keep inflation going in some places, even if it turns off elsewhere. In fact, even if it turns off “almost everywhere,” the tiny patches that continue to inflate will grow exponentially in volume. So the number of actual cubic centimeters in the inflating phase will grow without bound, leading to eternal inflation. Andrei Linde refers to such a picture as self-reproducing.

If inflation is eternal into the future, maybe you don’t need a Big Bang? In other words, maybe it’s eternal into the past, as well, and inflation has simply always been going on? Borde, Guth and Vilenkin proved a series of theorems purporting to argue against that possibility. More specifically, they show that a universe that has always been inflating (in the same direction) must have a singularity in the past.

But that’s okay. Most of us suffer under the vague impression — with our intuitions trained by classical general relativity and the innocent-sounding assumption that our local uniformity can be straightforwardly extrapolated across infinity — that the Big Bang singularity is a past boundary to the entire universe, one that must somehow be smoothed out to make sense of the pre-Bang universe. But the Bang isn’t all that different from future singularities, of the type we’re familiar with from black holes. We don’t really know what’s going on at black-hole singularities, either, but that doesn’t stop us from making sense of what happens from the outside. A black hole forms, settles down, Hawking-radiates, and eventually disappears entirely. Something quasi-singular goes on inside, but it’s just a passing phase, with the outside world going on its merry way.

The Big Bang could have very well been like that, but backwards in time. In other words, our observable patch of expanding universe could be some local region that has a singularity (or whatever quantum effects may resolve it) in the past, but is part of a larger space in which many past-going paths don’t hit that singularity.

The simplest way to make this work is if we are a baby universe. Like real-life babies, giving birth to universes is a painful and mysterious process. There was some early work on the idea by Farhi, Guth and Guven, as well as Fischler, Morgan and Polchinski, which has been followed up more recently by Aguirre and Johnson. The basic idea is that you have a background spacetime with small (or zero) vacuum energy, and a little sphere of high-density false vacuum. (The sphere could be constructed in your secret basement laboratory, or may just arise as a thermal fluctuation.) Now, if you’re not careful, the walls of the sphere will simply implode, leaving you with some harmless radiation. To prevent that from happening, you have two choices. One is that the size of the sphere is greater than the Hubble radius of your universe — in our case, more than ten billion light years across, so that’s not very realistic. The other is that your sphere is not simply embedded in the background, it’s connected to the rest of space by a “wormhole” geometry. Again, you could imagine making it that way through your wizardry in gravitational engineering, or you could wait for a quantum fluctuation. Truth is, we’re not very clear on how feasible such quantum fluctuations are, so there are no guarantees.

But if all those miracles occur, you’re all set. Your false-vacuum bubble can expand from a really tiny sphere to a huge inflating universe, eventually reheating and leading to something very much like the local universe we see around us today. From the outside, the walls of the bubble appear to collapse, leaving behind a black hole that will eventually evaporate away. So the baby universe, like so many callous children, is completely cut off from communication with its parent. (Perhaps “teenage universe” would be a more apt description.)

Everyone knows that I have a hidden agenda here, namely the arrow of time. The thing we are trying to explain is not “why was the early universe like that?”, but rather “why was the history of universe from one end of time to the other like that?” I would argue that any scenario that purports to explain the origin of the universe by simply invoking some special magic at early times, without explaining why they are so very different from late times, is completely sidestepping the real question. For example, while the cyclic-universe model is clever and interesting, it is about as hopeless as it is possible to be from the point of view of the arrow of time. In that model, if we knew the state of the universe to infinite precision and evolved it backwards in time using the laws of physics, we would discover that the current state (and the state at every other moment of time) is infinitely finely-tuned, to guarantee that the entropy will decrease monotonically forever into the past. That’s just asserting something, not explaining anything.

The baby-universe idea at least has the chance to give rise to a spontaneous violation of time-reversal symmetry and explain the arrow of time. If we start with empty space an evolve it forward, baby universes can (hypothetically) be born; but the same is true if we run it backwards. The increase of entropy doesn’t arise from a fine-tuning at one end of the universe’s history, it’s a natural consequence of the ability of the universe to always increase its entropy. We’re a long way from completely understanding such a picture; ultimately we’ll have to be talking about a Hilbert space of wavefunctions that involve an infinite number of disconnected components of spacetime, which has always been a tricky problem. But the increase of entropy is a fact of life, right here in front of our noses, that is telling us something deep about the universe on the very largest scales.

Update: On the same day I wrote this post, the cover story at New Scientist by David Shiga covers similar ground. Sadly, subscription-only, which is no way to run a magazine. The article also highlights the Banks-Fischler holographic cosmology proposal.

  1. Hi Administrator, great post.

    I think we can discount the singularity at the Big Bang
    The big bang itself would have proceeded from the singularity. This would be a true cyclical universe, unlike Turok’s et al. However the Timescale in which these cycles could or would occur, would be even beyond anything conceived to date – effectively making the universe for all intents and purposes eternal or infinite, since its beginning has ‘gone’ and there is no need of a foreseeable end – other than as an attempt to justify or verify this or that theory.

    But that is different from the existence of singularities in This Universe.
    The fact that we have Earth Gravity,
    The fact that we have the Sun’s Gravity
    and the fact that there is a gravitational field in the Milky Way on which the Sun orbits are enough to satisfy matter can become so compressed that it will reach blachole singularity properties.

    More like the galaxies themselves are the product of unlimited numbers of mini-bigbangs, thus they are the bubble universes or pocket universes that Susskind et al theorise about – but in This Universe.

  2. Sean,

    A very nice post on a very interesting question. The part where you sort out “big-bang” (well-tested evolution of an expanding homogeneous hot patch) from “big-bang” (an initial singularity) is important, as the two are often conflated even by people who know better.

    A couple of brief comments, hopefully more later. First, Borde, Guth & Vilenkin did *not* prove that eternal inflation has singularities to the past. As you know, most singularity theorems prove geodesic incompleteness, and this is the case here. What all of their theorems do are (a) write out a set of conditions which they consider to correspond to eternal inflation, then (b) show that the region in which these conditions hold is geodesically incomplete. This would indeed be consistent with eternal inflation “emerging from a primordial singularity”, but it is also consistent with eternal inflation just being grafted onto some spacetime region that is not eternally inflating by their definition. This is exactly what Steven & I did in various ways in our paper; and in most cases we argued that the ‘extra’ region was indeed eternally inflating, just not in accord with their criteria for eternal inflation.

    Second, I agree that it may be too constrictive to think of highly homogenous boundary conditions for the universe (then again, one person’s “fine-tuned” is another’s “simple and symmetric”), but a lot of the thinking of that paper — and yours with Chen — is independent of that. In particular, the notion that there is a surface of ‘minimal entropy’ (however large one would like to make it) from which the arrow of time points away in two directions, was treated in our paper using a homogeneous slice, but is clearly much more general, as you showed in your paper with Chen. And while I think you will agree that it is sensible, it appears to be largely new. Steven and I when writing our paper were amazed to find that aside from a few obscure references from decades ago, people discussing “time symmetric universes” essentially always talk about low entropy at the ‘beginning’ and ‘end’ of the universe, rather than high entropy at the ‘beginning’ and ‘end’, (i.e. low entropy in the ‘middle’). (Please correct me if I am wrong and if you know of other references.)

  3. This thread post reminded me of that other singularity: the Vinge-Kurzweil nodal point for human transformation. Sean mentions, in his opening paragraph, his prediction for post-2050 theoretical realizations. If there is any reliability to V-K’s prescient visions, then those realizations will be made by, or at least through, some transhuman conscious interactions. The computing power of self-programming AI’s are predicted to exceed human capacity by 2029, and “supposedly” by 2050 the capability of these “intelligences” will exceed the mental capacity of all humans put together. Thus, we could expect a transhuman consciousness to program and utilize vast networks of circuits (biological and mineral–meat and chip) to determine much of the probable theoretical validities regarding the cosmology of our Universe. I wonder though, just what manner of universes such consciousnesses perceive????

  4. Anthony, thanks for the clarification about the BGV results. About high entropy in the past and future, as far as I know you’re exactly right. Previous attempts to take seriously a time-symmetric cosmology have simply taken for granted that the past boundary condition is low-entropy, so the future one must be also (the Gold universe). As far as I know, your paper with Steven was the first in recent years to explore the obvious alternative.

  5. I know this isn’t on-topic, but you mentioned it here so I’m pouncing. Am I the only one irked by the use of “weightless” and “zero-gravity” wrt the Hawking story? There’s a big difference between being weightless and these Vomit Comet rides.

    Neither Hawking’s mass nor the Earth’s gravity field ceases to exist at 32k feet.


  6. Hi Sean,

    I love the classical interpretation of the big bang, because it falsifies all major religions while affirming theism. Thus, it makes nobody happy (except me).

    Next time I see you, let’s agree to bet an encyclopedia of religion that we won’t understand t

  7. oh… my less-than sign is being interpreted as an html tag.

    well… I mean tto say..

    Next time I see you, let’s agree to bet an encyclopedia of religion that we won’t understand t is less than or order of 0 in the next fifty years.

    My witty piths ruined by technology!


  8. There’s a big difference between being weightless and these Vomit Comet rides.

    Not really.

  9. Aaron,

    You still need to overcome gravitational forces to move the mass of your body. The weight of my left hand doesnt become zero simply because I’m embedded in a co-accelerating reference frame. In order to move my hand “up”, I still need to overcome its downward gravitational force. Furthermore, the gravitational effects on blood pressure and other physiological aspects don’t cease to exist. It may “look” like I’m weightless because of the reference frame, but it won’t “feel” like it.


  10. Pingback:

  11. I see the problem here. It’s not different than being in orbit. The inverse square law of gravity allows the stable Earth orbit. If you compute GM/r^2 for the ISS (~350km above sea level) you get g=~8.8m/s^2.

    Since gravity is required for orbit, you cannot be weightless and you cannot be in “zero-g”.

  12. Mollishka:

    Having spent far too many hours prowling through the Pharyngula comment threads, I’ve found that people who complain “why don’t the scientists talk about science, instead of religion or politics” are people offended that their own religious and/or political beliefs are being fisked. Of course, if you do write about some actual science, many of them will reject it from their brains post-haste because it conflicts with those self-same religious and political beliefs. Funny world, ain’t it?


    Now that the Horrendous Space Kablooie has come up, I wonder if you (or any of the other cosmologists in attendance) could take a look at the sidebar in Wikipedia’s article Bogdanov Affair. The text in the image caption (“Origin of the affair” section) appears to cover the same ground as this post, and AFAIK doesn’t have any egregious errors. It’d be nice to get a professional opinion.

  13. Pingback: Cosmic Variance: Speculating on the beginning of the universe « Identity Unknown

  14. Hands down this is one of the best posts I have ever read. Thanks a lot for posting it and for the links. I always like trying to find papers on this stuff.

    Sean, I know you like to post about more than science when you do you really hit the nail on the head. Thanks again.

  15. One of the questions associated with string theory/extra dimensions type approaches to origination is, why three large space dimensions, N =3? Normally, that is thought to derive from some process selection outcome, like Brandenberger and Vafa talking about number of dimensions constraining expansion of dimensions to larger size. Then there are Lisa Randall et al. with large hidden dimensions etc. Well, those approaches assume that spaces of other than three large dimensions are inherently self-consistent/non-contradictory (could exist without violating basic laws like conservation of energy, even if weird about signal transmission, atomic physics, bad for stable orbits and for organisms, etc.) There have been attempts in the past to find actual inconsistencies when N ≠ 3, which apparently aren’t convincing enough as actual prohibitions – however, it is now appreciated that 3-D space is indeed “special” in many ways.

    However, I have reviewed some apparently novel reasons for there being genuine self-contradiction where N ≠ 3 at Link, for anyone interested. If that pans out, then having N = 3 is not ultimately the result of a selection process that at least could have in principle turned out differently (and coincidentally, such processes could also still make N = 3 more likely, but they’d be “moot.”) It would instead be forced by a sort of logical constraint, if we think that the universe somehow tries to keep some basic principles going.

  16. EDT,
    This point is actually the cornerstone of general relativity, aka the equivalence principle. The effects of being in a gravitational field are indistinguishable from being in an accelerating frame. In the free-falling frame, the gravitational field has been locally transformed away. If you’re in the vomit comet which is in free fall and you let a ball go, you do not see it fall.

  17. Since gravity is required for orbit, you cannot be weightless and you cannot be in “zero-g”.

    …what V said.

    The reason Hawking was weightless was that he was falling at the same rate as the room around him.

    Another synonym for “zero-g” is “free fall,” and that term is perhaps a bit more accurate in its implications of what we really mean when we say weightless.

    You can always achieve weightlessness within a small “enough” region of spacetime. You don’t need to have “no gravitational fields” (or, perhaps to be more precise, the Riemann tensor all 0) in order to achieve the effects that we see as weightlessness and zero-g. You just have to be confined to a small enough region of spacetime that things freely falling don’t accelerate measurably relative to each other.


  18. I think it’s worth clarifying a little about “gravity vs. acceleration”. Uniform acceleration for an extended object is only equivalent to a uniform gravitational field. The gravitational field around the Earth is not uniform. Since the gravitational force is straight into the Earth, only the center of mass “feels” acceleration and gravity in an equivalent way. There are tidal forces on an object near a spherically symmetric object which will stretch it in the “vertical” direction and squish it in the “horizontal” directions. These are the forces that would mess you up if you were falling toward a dense object like a neutron star or black hole.

    Having just typed this, I see that Rob has made part of this point already.

  19. Sean,

    Is there a compelling reason why the cosmological constant must be positive? Instead of a false vacuum, perhaps we’re in an de Sitter bubble in an AdS true vacuum. Wouldn’t this make the birthing process less torturous?

  20. After reading a long and informative post that summarises our ignorance about the origin of the universe, I find it difficult to understand how any reasonable and rational human being can believe that…

    …we’ll understand what happened at the Big Bang within fifty years.

    But I find lots of things difficult to understand so I wouldn’t read anything into that.